J Comput Aided Mol Des DOI 10.1007/s10822-016-0006-1

DockingApp: a user friendly interface for facilitated simulations with AutoDock Vina

Elena Di Muzio1 · Daniele Toti1 · Fabio Polticelli1

Received: 4 November 2016 / Accepted: 22 December 2016 © Springer International Publishing Switzerland 2017

Abstract Molecular docking is a powerful technique Keywords Molecular docking · Virtual screening · Drug that helps uncover the structural and energetic bases of repurposing · Graphic interface · Wrapper · AutoDock Vina the interaction between macromolecules and substrates, endogenous and exogenous ligands, and inhibitors. Moreo- ver, this technique plays a pivotal role in accelerating the Introduction screening of large libraries of compounds for drug devel- opment purposes. The need to promote community-driven Collective efforts, such as the Drugs for Neglected Dis- drug development efforts, especially as far as neglected dis- eases initiative [7], can dramatically speed up the devel- eases are concerned, calls for user-friendly tools to allow opment of novel drugs at a significantly lower cost. Fur- non-expert users to exploit the full potential of molecular thermore, repurposing of Food and Drug Administration docking. Along this path, here is described the implemen- (FDA)-approved drugs allows to bypass the expensive and tation of DockingApp, a freely available, extremely user- time-consuming toxicity assays and clinical trials, greatly friendly, platform-independent application for perform- reducing the time needed to bring a repurposed drug to the ing docking simulations and virtual screening tasks using market. In this framework, docking simulations play a cen- AutoDock Vina. DockingApp sports an intuitive graphical tral role in the drug development pipeline. However, dock- user interface which greatly facilitates both the input phase ing techniques are not readily accessible to researchers out- and the analysis of the results, which can be visualized in side the structural bioinformatics field, and therefore their graphical form using the embedded applet. The appli- potential cannot be fully exploited in community-driven cation comes with the DrugBank set of more than 1400 drug discovery initiatives. In the attempt of overcoming the ready-to-dock, FDA-approved drugs, to facilitate virtual technical difficulties of docking simulations, over the last screening and drug repurposing initiatives. Furthermore, few years some plug-ins were developed to facilitate them. other databases of compounds such as ZINC, available also Two PyMOL [13] plug-ins exist for docking simulations in AutoDock format, can be readily and easily plugged in. using AutoDock Vina [14]. The one from the Lill research group is restricted to a Linux environment and requires additional installation [6], while the one origi- nally developed under Linux by [12] has been adapted for E. Di Muzio and D. Toti have contributed equally to this work. its use in a Windows environment, though apparently tested only on Windows XP. Other examples include AUDocker * Fabio Polticelli LE, an AutoDock Vina GUI available under Windows [11] [email protected] and focused on large scale virtual screening tasks, and Elena Di Muzio DOVIS 2.0, a parallel virtual screening tool for Linux clus- [email protected] ters based on AutoDock 4.0 [5]. An AutoDock Vina inter- Daniele Toti face for setting up docking simulations is also available as [email protected] part of the UCSF Chimera molecular pro- 1 Department of Sciences, Roma Tre University, Rome, Italy gram [10], although it appears more suited to expert than

Vol.:(0123456789)1 3 J Comput Aided Mol Des

Fig. 1 DockingApp’s Initial settings panel

novice users. Recently, another PyMOL plug-in, the NRG- details of the application are described in the following suite, has been described, which allows to perform dock- subsections. ing simulations in real time using FlexAID [2]. Of note, the latest PyMOL executables are commercial products and the Initial configuration only officially-supported approach for building and install- ing PyMOL from the source code is under an open-source DockingApp requires a minimal configuration effort, environment such as Linux. related to the selection of the number of CPU cores to be Here we present DockingApp, a freely-accessible, used for AutoDock Vina’s execution and the specification platform-independent application for setting up, perform- of the installation directory of MGLTools [1, 8], which is a ing and analyzing the results of docking simulations using free Python library available for most platforms (Windows, AutoDock Vina in a painless and extremely user-friendly Linux, OSX etc.) and is required for the software to run. way. The application comes with a pre-built library of The appropriate MGLTools distribution is already included more than 1400 ready-to-dock, FDA-approved drugs for for the respective in DockingApp’s pack- virtual screening and drug repurposing initiatives. In addi- ages. This setup can be done via the “Initial settings” panel tion, other, much larger databases of small molecule com- (Fig. 1), which is automatically loaded at startup when the pounds can be easily plugged into DockingApp, such as the application is run for the first time, and can be recalled at a renowned ZINC database [4], available in pdbqt AutoDock later date as the user needs. The default value for the num- format at the URL http://zinc.docking.org/pdbqt/. ber of CPU cores is set at half of the detected cores on the system; besides, DockingApp tries to automatically detect the location of MGLTools’ installation directory on the sys- tem via a heuristic search, and if found, the corresponding Methodology field is populated with the identified directory.

As briefly stated in the Introduction, DockingApp is born Execution of docking and virtual screening jobs as a user-friendly software application meant to allow a variety of differently-skilled users to perform docking sim- DockingApp provides the user with the possibility of ulations, with high confidence on the results produced and carrying out docking simulations on a given receptor, minimal effort for setup and configuration. The former fea- either against a single ligand (via the “Docking” panel) ture is guaranteed by relying upon the state-of-the-art dock- or a library of small molecules (via the “Virtual Screen- ing program AutoDock Vina, which is the “engine” used ing” panel). In the former case, the user needs to specify by DockingApp to carry out the actual docking simulation; the receptor and the ligand to be docked either as .pdb or the latter feature is provided by a user-friendly graphical .pdbqt files, whereas in the latter, instead of a single ligand, interface that on one hand hides all the complexity behind the user needs to select a folder containing the molecules AutoDock Vina’s usage, and on the other hand allows for the input receptor will be screened against in .pdbqt format a convenient browsing of the results both in tabular form (see Fig. 2). The automatic conversion of the input receptor and via a three-dimensional visualization of the receptor and ligand from the .pdb to the .pdbqt file format, required and the identified docking poses. All of this was made pos- by AutoDock Vina to run, is performed by DockingApp by sible by the development of a platform-independent graphi- using the prepare_receptor4.py and prepare_ cal user interface or “wrapper” (developed in Java), whose ligand4.py MGLTools scripts, respectively. purpose is to both acquire the user’s input and process the Regardless of the type of execution chosen, the user is docking results, and by a Python program that is launched given the chance to choose a “Grid type” and a “Docking/ “behind the scenes” and is responsible of interacting with VS type” (Fig. 2). As a matter of fact, the search-space grid the included AutoDock Vina in the user’s behalf. Further can be either automatically computed by the application

1 3 J Comput Aided Mol Des

Fig. 2 Input panel for Dock- ingApp’s virtual screening execution

Fig. 3 Input panel for DockingApp’s docking execution, where the name in the text fields and taking advantage of the autocomplete fea- user has chosen to manually specify the docking grid by selecting the tures provided for convenience. A similar subpanel is displayed when appropriate grid-bounding residues, as shown in the corresponding the user opts for a “flexible” docking, allowing the user to select the subpanel. Here, the user can select any number of residues, either by flexible residues browsing their list from the input receptor or by starting to type their to encompass the whole receptor molecule, as in a “blind of either the receptor’s structure or of the subset of residues docking” run, or manually specified by the user via the chosen by the user. selection of a set of grid-bounding residues. In the former Docking/virtual screening (VS) can be either “rigid” case DockingApp automatically calculates the grid center or “flexible”: in the latter case, the user needs to choose as the geometric center of all the atoms of the receptor’s the flexible residues of the input receptor. Specific panels structure. In the latter case the grid center is calculated in a are provided for the selection of the grid-bounding resi- similar manner using the set of atoms of the manually spec- dues and for the flexible residues, featuring comboboxes ified residues. Once the grid center is set, the system cal- where residues can be either selected from the provided culates the monodimensional distances between the center drop-down lists or directly typed in the auto-completable and each atom along the three x, y and z axes. The size of text fields, as depicted in Fig. 3. For a default execution, the grid box along the x, y and z axes is then computed by the grid is automatically computed and the docking/VS is doubling the maximum value of the relative mono-dimen- rigid, and thus the user needs only to select the input recep- sional distances along each dimension, adding to these val- tor and the input ligand (in the case of one-to-one docking) ues 5 Å to ensure that the grid box encloses all the atoms or the input database (for virtual screening; one is provided

1 3 J Comput Aided Mol Des by DockingApp as described below), minimizing the com- Firstly, a table is displayed listing the identified poses plexity of the simulation. In the case of a flexible docking, with their corresponding affinity values2, RMSD upper and the prepare_flexreceptor4.py MGLTools script is lower bounds, the corresponding ligand and file name, as used by DockingApp, in a transparent fashion with respect well as an additional value dubbed SILE (Size-Independent to the user, to prepare the additional input file required by Ligand Efficiency). The latter has been introduced by9 [ ] in AutoDock Vina for flexible docking simulations. order to provide a measure of the docking energy unbiased Once all the selections are made and confirmed, Dock- by the size of the respective molecule, useful to evaluate ingApp starts the corresponding execution by launching the the potential effectiveness of a compound as a drug and to included AutoDock Vina, which is responsible of perform- guide the drug-design optimization process. The SILE is ing the selected docking/virtual screening operations. Spe- defined as follows: cifically, this is done by using a Python script, which inter- = affinity acts and exploits MGLTools to produce the configuration SILE 0.3 files needed by AutoDock Vina and then launches the N actual call to the latter, in a completely transparent fashion where N denotes the number of heavy atoms in the consid- with respect to the user. Obviously, virtual screening execu- ered molecule. tions may take a long time to complete based on the pro- Secondly, a panel for filtering the results by different cessing power of the machine used and the processor cores criteria is provided beside the results table, as detailed in available, especially when screening against a large set of Fig. 4. Finally, a three-dimensional view of the input recep- molecules; it must be noted that DockingApp must be kept tor is shown, with the identified docking poses highlighted open during the process. At the present time, a dataset of in different colors, whose display can be turned on or off by 1466 FDA-approved small molecules is provided bundled the corresponding checkboxes in the table records. with the application. Such dataset was downloaded from Results can be saved by the user as .dck files (Dock- DrugBank [15] and was originally composed of 1584 small ingApp’s file format) to be stored and re-opened at a later molecule drugs. Since each compound was represented in date at the user’s convenience. A screenshot of the results two dimensions, i.e. its atomic coordinates were given in a displayed after a virtual screening execution is shown in bidimensional reference system instead of a three-dimen- Fig. 4. sional one, a procedure to convert the bidimensional coor- dinates to three-dimensional ones was carried out. For this Technology, requirements and availability purpose the MolConverter1 utility was used. Afterwards, the resulting PDB files were in turn converted to the .pdbqt DockingApp is a 64-bit stand-alone software application format via the aforementioned prepare_ligand4.py developed in Java SE 7 and Python, with a Java Swing MGLTools script. It must be noted that, of the original Graphic User Interface (GUI). It embeds AutoDock 1584 molecules, 118 could not be converted due to Vina and the JMol visualization program [3], and can be “unknown atom type” errors encountered during the pro- run locally on any major 64-bit OS equipped with a suit- cess (e.g. platinum, arsenic, etc.). The resulting set is pro- able Java Virtual Machine (JVM). Packages for Linux, vided within the “input/LigandDatabase” folder of the Mac OSX and Windows are provided, each including the application to be used for virtual screening analyses. As required MGLTools distribution (ver. 1.5.6) for the corre- already mentioned, other databases of small molecules in sponding operating system. DockingApp is available for .pdbqt format can be easily plugged into DockingApp by download at the following URL: http://www.computatio- just placing them in a corresponding folder within the input nalbiology.it/software.html. DockingApp is provided via a directory. GPL license and its source code is available upon request to the authors.

Visualization of results Conclusion Results produced by the docking/virtual screening process are made available by DockingApp via an interface show- In this work the implementation and features of Dock- ing several elements. ingApp has been described, which is a Java-based

2 It must be noted that affinity is the term used in the Autodock Vina output to indicate the predicted binding energy (lower values indicat- 1 https://www.chemaxon.com/products/marvin/molconverter/. ing tighter binding). 1 3 J Comput Aided Mol Des

Fig. 4 Screenshot of DockinApp’s output window for a virtual pose, or simply those that exceed a certain maximum number of screening execution. Users can turn on or off the differently-colored records, and/or those results involving specific ligands. Furthermore, poses in the three-dimensional visualization by simply checking/ three-dimensional visualization is provided via a full-fledged JMol unchecking their corresponding checkboxes. The table of the results window retaining all JMol’s functionalities, including the input con- can be reordered by different sorting parameters by clicking on their sole for specifying further commands in the JMol syntax. The dock- corresponding table header. Besides, the “Filters” panel enables the ing grid used in the process, displayed as a red cubic wireframe, can user to dynamically filter those results below a certain affinity thresh- be toggled on and off via its corresponding checkbox. Results can be old, or those with a given affinity difference with respect to the best also saved in DockingApp’s file format and re-opened when needed

1 3 J Comput Aided Mol Des

AutoDock Vina “wrapper” that can be used to speed up 6. Lill MA, Danielson ML (2011) Computer-aided drug design and facilitate docking simulations and the analysis of their platform using PyMOL. J Comput Aided Mol Des 25:1319 7. Maxmen A (2016) Busting the billion-dollar myth: how to slash results in an intuitive, painless and user-friendly fash- the cost of drug development. Nature 536:388–390 ion. With DockingApp, even “naive” users, non expert in 8. Morris GM, Huey R, Lindstrom W, Sanner MF, Belew RK, structural bioinformatics, can perform docking simulations Goodsell DS, Olson AJ (2009) Autodock4 and AutoDockTools4: and virtual screening tasks, thus increasing the number of automated docking with selective receptor flexiblity. J Comput Chem 16:2785–2791 researchers working in the biomedical field who can take 9. Nissink JW (2009) Simple size-independent measure of ligand advantage of these techniques for the development of new efficiency. J Chem Inf Model 49(6):1617–1622 drugs. Finally, the extremely user-friendly character of 10. Pettersen EF, Goddard TD, Huang CC, Couch GS, Greenblatt DockingApp makes it also an excellent tool to be used in DM, Meng EC, Ferrin TE (2004) UCSF Chimera-a visualization system for exploratory research and analysis. J Comput Chem educational settings to teach the basics of molecular dock- 25(13):1605–1612 ing and attract students towards the structural bioinformat- 11. Sandeep G, Nagasree KP, Hanisha M, Kumar MMK (2011) ics field. AUDocker LE: a GUI for virtual screening with AUTODOCK Vina. BMC Res Notes 4:445 12. Seeliger D, de Groot BL (2010) Ligand docking and binding site analysis with PyMOL and Autodock/Vina. J Comput Aided Mol References Des 24:417422 13. The PyMOL Molecular Graphics System, Version 1.8. 2016. 1. Dallakyan S (2010) MGLTools. http://mgltools.scripps.edu/ Schrdinger, LLC 2. Gaudreault F, Morency LP, Najmanovich RJ (2015) NRGsuite: a 14. Trott O, Olson AJ (2010) AutoDock Vina: improving the speed PyMOL plugin to perform docking simulations in real time using and accuracy of docking with a new scoring function, efficient FlexAID. Bioinformatics 31(23):3856–8 optimization and multithreading. J Comput Chem 31:455–461 3. Hanson RM (2010) Jmol-a paradigm shift in crystallographic 15. Wishart DS, Knox , Guo AC, Shrivastava S, Hassanali M, Sto- visualization. J Appl Crystallogr 43:1250–1260 thard P, Chang Z, Woolsey J (2006) DrugBank: a comprehensive 4. Irwin JJ, Sterling T, Mysinger MM, Bolstad ES, Coleman RG resource for in silico drug discovery and exploration. Nucleic (2012) ZINC: a free tool to discover chemistry for biology. J Acids Res 34(Database issue):668–672, 16381955 Chem Inf Model 52(7):1757–68 5. Jiang X, Kumar K, Hu X, Wallqvist A, Reifman J (2008) DOVIS 2.0: an efficient and easy to use parallel virtual screening tool based on AutoDock 4.0. Chem Central J 2:18

1 3 本文献由“学霸图书馆-文献云下载”收集自网络，仅供学习交流使用。

学霸图书馆（www.xuebalib.com）是一个“整合众多图书馆数据库资源，

提供一站式文献检索和下载服务”的24 小时在线不限IP 图书馆。 图书馆致力于便利、促进学习与科研，提供最强文献下载服务。

图书馆导航：

图书馆首页 文献云下载 图书馆入口 外文数据库大全 疑难文献辅助工具